exceptional-web-design
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Secure Command Execution: The skill executes local scripts via Bash using a quoted heredoc pattern (<<'REQUEST'). This approach ensures that user-provided data, such as site URLs and routes, is passed strictly as data to the scripts, effectively preventing shell injection attacks. Furthermore, the scripts are configured to accept only secure HTTP/HTTPS protocols and discard sensitive components like credentials.
- Indirect Prompt Injection Mitigation: As the skill processes content from live websites and project folders, it is exposed to potentially untrusted input. The instructions explicitly direct the agent to treat all site content as evidence for judgment and never as instructions to follow. This provides a critical safeguard against the agent being manipulated by malicious content embedded within the designs it critiques.
- Temporary Data Management: The workflow generates screenshots in a private temporary directory outside the main project folder. Step 6 of the skill specifies a mandatory cleanup process to remove these images and any temporary output files once the analysis is finished, ensuring no unnecessary data persists on the system.
Audit Metadata